The Impact of Cognitive Anchors on Information-Sharing Behavior

Abstract

In computer-supported information exchange, people frequently tend to withhold their own information and free-ride on the others' contributions. In doing so, they save costs (time and effort) and maximize their own benefit. However, if everyone behaved in this way, there would be no information sharing at all. In this experiment, we tested if the presentation of a random number could serve as a cognitive anchor and influence the amount of shared information. The experimental setting had all the features of an information-exchange dilemma. Before participants could share information, a random generator presented a random number. It was found that this number served as a cognitive anchor and influenced both the participants' behavioral intentions and their actual behavior. Particularly, the high anchor increased cooperation, even though enhanced cooperation was obviously detrimental to the individual's own benefit.

Introduction

Today many processes of information sharing and decision making are carried out through the Internet or other computer networks.^1–4 This allows people to access all kinds of information from all over the world. Very often, this information exchange uses shared databases.^5,6 Anyone who has access to the Internet may enter information into such a database and retrieve information that was provided by others. However, even though information sharing would seem to be highly beneficial, it can often be observed that the exchange of information does not flow as smoothly as expected. Frequently, people are reluctant to share their information. This problem has been attributed by various authors to the fact that this type of information sharing may have the typical characteristics of a social dilemma.^7–12 An information-exchange dilemma will occur in those situations in which everyone has access to everyone else's contributions, but in which it is also easy for individuals to use the other participants' information without ever contributing any information themselves.¹³ So, it would be the most profitable strategy from an individual's point of view to keep one's own information to oneself and still use that provided by the others.¹¹ However, if everybody behaved in this way, then the database would be empty, and no one in the group could benefit. The fact that thriving cooperation has been shown to be an important prerequisite for the success of teams and organizations in various contexts^14–16 justifies research on computer-supported cooperation. Such research should focus on identifying factors that influence people's willingness to collaborate with each other. The potential for application of this research makes it very important. Thus, this article aims at experimentally examining a factor that might subtly sustain a cooperative behavior in computer-supported groups, namely the use of cognitive anchors while sharing information on the Internet.

In the following section, we will present recent findings of previous studies that concern what impact behavioral guidelines have on people's cooperative behavior in computer-supported information-exchange situations. We will explain why the underlying theoretical mechanism did not become clear as a causal factor in the results of these previous studies. The current article explicitly aims at filling this theory gap: we will illustrate in which way an anchoring effect may be the explanation for the earlier findings. This presentation of theory will result in two hypotheses, which state that cognitive anchors are able to influence people's willingness to cooperate in information-sharing situations. We will then explain the method we used to test these hypotheses. Finally, we will present our findings and conclude with a discussion of these findings and their implications for further research and application.

Impact of Behavioral Guidelines

Cress and Kimmerle¹⁷ have examined an information-exchange situation in which the participants worked in six-person groups in the same room. They were made to believe that their computers were connected through a shared database, but in reality, the behavior of the others was simulated by a computer program. The setting created a social dilemma in which participants had the choice whether to share information or not. The participants' task was constructed in such a way that they needed certain pieces of information that the other participants had worked out. However, contributing information meant losing time. As a consequence, an individual's best choice was not to share their own information items, but to use, those that had been contributed to the shared database by others. To examine the effect of guidelines on people's behavioral intentions and actual behavior, the participants in this study (Study 1 in Cress and Kimmerle¹⁷) received either the recommendation to contribute eight pieces of information (high-level guideline), to contribute three pieces of information (low-level guideline), or (in the control group) they did not receive any recommendation at all. Cress and Kimmerle¹⁷ had assumed that with a high-level guideline the participants' willingness to share information would be higher than with a low-level guideline or in the control group. The underlying theoretical consideration was that guidelines would be effective due to their character as cognitive anchors in the sense of Tversky and Kahneman.^18–20

In Tversky and Kahneman's experiments,²⁰ the assignment had been to give quantitative estimates. Participants had to assess, for example, the percentage of African countries in the United Nation Organization. Before they made their assessment, they watched a wheel of fortune that was spun in their presence. Then, they had to estimate if the number in demand (percentage of African countries in the UN) was higher or lower than the number determined by the wheel of fortune (this random number provided by the wheel of fortune will be called the anchor value in the remainder of this article). After this comparison, estimation they had to state the exact number (the actual percentage of African countries). This assessment of the exact number is called the absolute estimation. Tversky and Kahneman²⁰ found that absolute estimations differed from each other, depending on the anchor value. If participants were provided with a high random number, then their estimations turned out to be significantly higher than with a low anchor value. This result was labeled the anchoring effect. On the basis of these findings, Cress and Kimmerle¹⁷ suggested that the fact that behavioral guidelines did have an impact on people's behavior might be due to such an anchoring effect as well. They argued that people in an information-exchange dilemma find themselves in a situation of uncertainty, so they are inclined to look for cues that might give them a hint about appropriate behavior. This line of argument represents an extension to the research by Tversky and Kahneman, in that the cognitive anchor should have an impact not only on assessments, but also on actual behavior, and indeed, Cress and Kimmerle¹⁷ found that guidelines had such effects on behavior as had been expected.

Kimmerle et al.²¹ have transferred this experimental setting into an online situation. For this purpose, an online game was devised in which the participants played the role of detectives. To be able to solve their criminal cases with a higher probability, they had, again, the opportunity to share individual information items or not. It was found that behavioral guidelines had a significant impact on the number of contributions that participants provided, in such a way that high-level guidelines led to significantly more contributions than low-level guidelines or no guidelines.

However, in all these studies, the underlying mechanism did not become clear as a causal factor in the results. It is possible that what is reported here was a cognitive effect in terms of anchoring. It is also plausible however that people adhered to behavioral recommendations, because they regarded them as some kind of prescriptive social norm²²; that is, they conformed to guidelines, because they believed that this was what other people expected of them—since a recommendation may imply some kind of behavioral instruction. In the study presented here, we wanted to exclude social norms as an explanation for the impact of guidelines. For this purpose, our experiment was constructed in such a way that if people were influenced by numerical standards, this could only be explained validly by cognitive anchoring. Therefore, we did not use behavioral recommendations here, but instead chose a procedure that was geared to the classical method of Tversky and Kahneman and applied random numbers as cognitive anchors. We are not denying of course that social norms might have an impact on people's behavior in an information-exchange dilemma. However, what the experiment will demonstrate explicitly is that an anchor that is not a behavioral recommendation and that operates at the cognitive level is able to elicit a modification of a cooperative behavior, even in a situation in which enhanced cooperation is obviously detrimental to the individual.

Hypothesis 1: People will adjust their behavioral intention to a randomly presented number (anchor value). With a high random number, they will express the intention to share a higher number of information items than with a low random number.

Hypothesis 2: People will adjust their actual behavior to a randomly presented number. With a high random number, they will share a higher number of information items than with a low random number, even though this behavior is against their own benefit.

Methods

Participants

Seventy-six participants took part in this study. One participant had to be excluded due to a wrong answer in the manipulation check (see below). Out of the remaining 75 participants, 41 were women, and 32 were men, and two did not indicate their gender. Their mean age was 27.23 years (SD=11.51). Participants were randomly assigned to one of the three experimental conditions with either a high-, a low-, or no-anchor value. Twenty-six participants were assigned to the high-, 23 to the low-, and 26 to the no-anchor condition.

Experimental setting

The experiment was carried out as an online study on the Internet. The basis for this experiment was the experimental setting used by Kimmerle et al.²¹ (cf. also Wodzicki et al.²³). This setting was adapted to the needs of the current study. The experiment was conceptualized as an online game in which the participants played the roles of private investigators who had to solve two criminal cases. To be able to solve these cases, they needed certain items of information. The participants were told that they were one of six detectives who had to catch a thief; that is, participants were led to believe that they would solve the cases together with five other participants who were taking part simultaneously (but in reality, these other participants' behavior was simulated by a computer program). According to the story presented in the game, a thief had stolen art treasures from a museum. The participants had to identify the thief and trace the hiding places of the treasures. Participants were informed that they had the chance of winning vouchers for an Internet store (25 € each). The more thieves they were able to identify and the more hiding places they found, the higher their chances were to win the voucher. In each of the two cases, all participants started with nine clues that hinted hiding places of the treasures as well as the identity of the thief. However, these nine clues were not sufficient to solve the case. Therefore, the participants had to cooperate with each other. They could contribute any proportion of their clues to a shared database that could be accessed by the other detectives. In this way, a game was created that had the payoff structure of a social dilemma: the more pieces of information participants contributed to the database, the lower their chance was of winning the voucher; but the fewer pieces of information the database contained, the smaller an individual participant's chance was of winning the voucher.

For manipulating the value of the anchor, a random generator (with numbers from zero to nine) was used. The participants saw a text on the computer screen that requested them to use the random generator to determine their personal number. They were told that this was their number for logging in, and that they should keep that number in mind. The random generator had the appearance of a gambling machine, and the participants were able to stop its fast rotation by clicking a button. In the low- and the high-anchor condition, the random generator always indicated 2 and 7, respectively. In the control condition, no such random generator appeared. Subsequently, in the experimental conditions, the participants had to answer two questions—analogously to the comparison and absolute estimation in Tversky and Kahneman's studies.²⁰ The first question was if they would contribute more or fewer than two/seven clues to the shared database. Then, they were asked how many clues they intended to contribute in absolute numbers; here, they had to enter the respective digit (0–9). In the control condition, only the second question appeared.

Measures

As a manipulation check, participants in the experimental conditions had to enter their personal number before each of the two trials. All participants had to fill in a pre-experimental questionnaire concerning demographic data. The dependent variables were people's intention to share information items measured via their absolute estimation and the number of shared information items, that is, people's behavior.

Procedure

Participants used their own computers and the Internet for the experiment. They were informed that the study gave them the opportunity to win vouchers for an Internet store. Before the actual game started, they had to complete the pre-experimental questionnaire. Then, the rules and sequence of events in the game were presented, and the payoffs of the information-exchange dilemma explained in detail. After that, the participants in the experimental conditions were assigned their random number, asked about their intended number of contributions (in a comparison [simply more or fewer] and in an absolute [how many exactly] format), and then entered their personal number as a manipulation check. In the control condition, participants were only asked what their absolute number of contributions would be. Then, the first trial of the game started, and the participants had the opportunity to contribute their information items to and retrieve other information items from the database to try to solve the first case. This was followed by the second trial, which was identical to the first one. Finally, the participants were debriefed.

Results

For testing the two hypotheses, we conducted analysis of variances (ANOVAs) with anchor as a fixed factor for comparing the experimental conditions for each dependent variable. The ANOVA supported Hypothesis 1 and revealed a significant main effect for intention, F(2, 72)=9.70, p<0.001, η_p²=0.21. Post hoc tests (Fisher's least significant difference [LSD] test) showed that a high anchor value led to a behavioral intention that was significantly higher than in the condition with the low anchor (M_high=6.92, SD=2.45 vs. M_low=4.00, SD=2.76, p<0.001) and the condition with no anchor (M_high=6.92, SD=2.45 vs. M_no=4.69, SD=2.15, p<0.01). The conditions with a low anchor and without an anchor did not differ with respect to people's behavioral intention, p=0.33, ns.

The ANOVA also supported Hypothesis 2 and revealed a significant main effect for behavior, F(2, 72)=4.69, p=0.01, η_p²=0.12. Post hoc tests (LSD) showed that a high anchor led to significantly more contributions than a low anchor (M_high=4.90, SD=3.08 vs. M_low=2.70, SD=2.84, p<0.01) and no anchor (M_high=4.90, SD=3.08 vs. M_no=3.38, SD=1.70, p=0.04). The conditions with a low anchor and without an anchor did not differ with respect to behavior, p=0.36, ns.

Discussion

In situations of database-mediated information exchange, people tend to share only a few information items, but still use information that other people have contributed.²⁴ This behavior results from the social dilemma that is present in such situations.^25,26 Recent studies^17,21 have shown that guidelines that were presented as behavioral recommendations were able to influence people's information-sharing behavior. In the literature, two competing explanations are cited. It has been argued that adhering to such recommendations is an effect of cognitive anchoring. However, it is also possible that behavioral guidelines are effective, because people perceive the recommendations as a prescriptive social norm that they had to comply with.

Consequently, the study that is described in the current article aimed at creating a situation in which a prescriptive social norm as an explanation could be ruled out: it was examined whether a random number (not a behavioral recommendation) provided by a random generator was able to influence people's information-exchange behavior. The results showed that a high cognitive anchor led to a more pronounced intention to share a high number of information items than a low anchor or a situation without an anchor. So far, this corresponds to the classical anchoring effect, as it was triggered by our treatment: the cognitive anchor influenced people's behavioral intention according to their answers in the estimation task. However, and this is remarkable, analogous effects could also be observed in their actual behavior: a high anchor was able to increase the participants' contribution rate. There is an obvious explanation for the fact that it was particularly the high anchor that modified people's behavior, while the low anchor did not have a significant effect compared to the control group: a low anchor, used as a behavioral guideline by the participants (here: to share two information items), may largely correspond to the behavior that people would have exhibited anyway (as in the control group), whereas the high anchor elicited an improvement of cooperation. From the perspective of application, this is a positive finding, since impairment of cooperative behavior by a low anchor is hardly ever intended in real life. Enhancement of cooperation is more likely to be desired in a variety of situations.

The impact of the random number on behavior in the information-exchange dilemma is noteworthy: so far, only effects on estimations and judgments, but not on actual conduct, have been reported in pertinent literature. Moreover, the experimental setting that was used in this study represents a particularly strict test: when people adjusted their behavior due to a high anchor, when they made more contributions than with a low or without any anchor, then they acted explicitly against their own interests, because they impaired their chances of winning a valuable voucher. The fact that the participants with a high anchor were still more cooperative is a remarkable demonstration of the power of the anchoring effect in an information-exchange dilemma. Apparently, participants in the high-anchor condition did not feel that their behavior was irrational, because they had the opinion that sharing many pieces of information was a particularly appropriate behavior.

Our findings are relevant both to basic and applied research: on one hand, the study makes an important theoretical contribution toward understanding underlying processes of people's information-sharing behavior. From a perspective of application, on the other hand, it is relevant that the study demonstrates an option for alleviating some problems that occur in information-exchange dilemmas. Therefore, we hope that our study will stimulate further research in both areas: examining and analyzing the mode of operation of the anchoring effect in various contexts, and developing simple and pragmatic solutions for situations of computer-mediated information sharing.

Future studies should consider how persistent the anchoring effect is over time. Previous studies have found that the effect is rather durable.²⁷ However, as yet, it is unclear whether this is also true of the impact on information-sharing behavior on the Internet. In this context, theoretically sound considerations should be advanced as to how this effect could be supported in the practice. An example would be to repeat the comparison and absolute estimation tasks throughout the information-sharing process. This would be an important endeavor from the perspective of potential application. From a theoretical point of view, it will be interesting for future studies to examine in what way cognitive anchors are mentally represented to allow them to have an impact on behavior. One may assume that, for example, anchors make behavior-guiding cognitions simply more easily accessible.²⁸ Whether this is indeed the case is an important question for further studies.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

Chen

. Auction fever: exploring informational social influences on bidder choices. Cyberpsychology, Behavior, and Social Networking, 2011; 14:411–416.

LaRose

, Kim

. Share, steal or buy? A social cognitive perspective of music downloading. Cyberpsychology and Behavior, 2007; 10:267–277.

Lin

, Wang

. A decision-making process model of young online shoppers. Cyberpsychology and Behavior, 2008; 11:267–277.

Stefanone

, Lackaff

, Rosen

. Contingencies of self-worth and social-networking-site behavior. Cyberpsychology, Behavior, and Social Networking, 2011; 14:41–49.

Jian

, Jeffres

. Understanding employees' willingness to contribute to shared electronic databases: a three-dimensional framework. Communication Research, 2006; 33:242–261.

Kalman

, Monge

, Fulk

et al. Motivations to resolve communication dilemmas in database-mediated collaboration. Communication Research, 2002; 29:125–154.

Bonacich

. Communication dilemmas in social networks: an experimental study. American Sociological Review, 1990; 55:448–459.

Cabrera

, Cabrera

. Knowledge-sharing dilemmas. Organization Studies, 2002; 23:687–710.

Cheshire

. Selective incentives and generalized information exchange. Social Psychology Quarterly, 2007; 70:82–100.

10.

Cress

, Kimmerle

. Endowment heterogeneity and identifiability in the information-exchange dilemma. Computers in Human Behavior, 2008; 24:862–874.

11.

Cress

, Kimmerle

, Hesse

. Impact of temporal extension, synchronicity, and group size on computer-supported information exchange. Computers in Human Behavior, 2009; 25:731–737.

12.

Hollingshead

, Fulk

, Monge

. 2002. Fostering intranet knowledge sharing: an integration of transactive memory and public goods approaches. Hinds

, Kiesler

. Distributed work. Cambridge, MA: MIT Press, 335–355.

13.

Kimmerle

, Cress

. Group awareness and self-presentation in computer-supported information exchange. International Journal of Computer-Supported Collaborative Learning, 2008; 3:85–97.

14.

Chang

C-W

, Wu

C-R

, Liao

C-C

. The virtual cooperation platform in enterprise and supplier cooperation models. Cyberpsychology, Behavior, and Social Networking, 2010; 13:473–475.

15.

C-R

, Chang

C-W

, Chen

C-C

. Evaluate collaborative design semiconductor industry statistical process control based on GSDM. Journal of Grey System, 2010; 22:275–280.

16.

C-R

, Chang

C-W

, Liao

C-C

. Selection of collaborative technology used by software contractors. Journal of Testing and Evaluation, 2010; 38:731–737.

17.

Cress

, Kimmerle

. Guidelines and feedback in information exchange: the impact of behavioral anchors and descriptive norms in a social dilemma. Group Dynamics: Theory, Research, and Practice, 2007; 11:42–53.

18.

Kahneman

. Reference points, anchors, norms, and mixed feelings. Organizational Behavior and Human Decision Processes, 1992; 51:296–312.

19.

Kahneman

, Slovic

, Tversky

. 1982. Judgment under uncertainty: heuristics and biases. New York: Cambridge University Press.

20.

Tversky

, Kahneman

. Judgment under uncertainty: heuristics and biases. Science, 1974; 185:1124–1131.

21.

Kimmerle

, Wodzicki

, Jarodzka

et al. Value of information, behavioral guidelines, and social value orientation in an information-exchange dilemma. Group Dynamics: Theory, Research, and Practice, 2011; 15:173–186.

22.

Reno

, Cialdini

, Kallgren

. The transsituational influence of social norms. Journal of Personality and Social Psychology, 1993; 64:104–112.

23.

Wodzicki

, Schwämmlein

, Cress

et al. Does the type of anonymity matter? The impact of visualization on information sharing in online groups. Cyberpsychology, Behavior, and Social Networking, 2011; 14:157–160.

24.

Kimmerle

, Cress

, Hesse

. An interactional perspective on group awareness: alleviating the information-exchange dilemma (for everybody?) International Journal of Human-Computer Studies, 2007; 65:899–910.

25.

Cress

, Kimmerle

, Hesse

. Information exchange with shared databases as a social dilemma: the effect of metaknowledge, bonus systems, and costs. Communication Research, 2006; 33:370–390.

26.

Kimmerle

, Cress

. Visualization of group members' participation: how information presentation formats support information exchange. Social Science Computer Review, 2009; 27:243–261.

27.

Mussweiler

. The durability of anchoring effects. European Journal of Social Psychology, 2001; 31:431–442.

28.

Mussweiler

, Strack

. Hypothesis-consistent testing and semantic priming in the anchoring paradigm: a selective accessibility model. Journal of Experimental Social Psychology, 1999; 35:136–164.